The world didn’t stop lending to America. It stopped lending long.

• Nobody is dumping Treasuries. Foreigners bought $207.1 billion of long-term US securities in June 2026. What changed is the buyer: private money does the work now, central banks stopped being the marginal bid years ago, and private money charges for duration.
• With net interest at $1.04 trillion and a record 18.6% of federal revenue, the Treasury is shortening the debt rather than paying the long price. It is also lending Japan dollars against its Treasuries through the Fed’s FIMA window, so the bonds never have to be sold and the sale never shows up in the data.
• It isn’t working yet. On 9 September the Treasury confirmed a buyback of up to $6 billion and the ten-year closed at a near three-year high the same day. The engineered demand is real and still small, and the adjustment is showing up in what patient money buys instead: 289 tonnes of central-bank gold in a single quarter, bought into a falling price.

On 4 September the president of the United States threatened to stop trading with every country that runs a surplus against it, unless the Federal Reserve cut interest rates. He posted that an embargo would be “BETTER THAN TARIFFS”, and told reporters he was under no obligation to allow trade at all.

The reading that spread from there was tidy. America owes more than it can afford, the countries it owes are heading for the exit, and the threats are what a borrower does when the lenders stop showing up.

Half of that is right. The fiscal arithmetic is genuinely bad and it is getting worse on schedule. The other half, the part about lenders leaving, does not survive contact with the filings. Foreigners bought $207.1 billion of long-term US securities in June. What changed is who is buying, and for how long.

Split the foreign bid into its two halves and the story separates cleanly.

In June, private foreign investors bought $169.8 billion of long-term US securities. Foreign official institutions, meaning central banks and sovereign funds, bought $37.3 billion. In March the official sector was a net seller of $14.9 billion while private investors bought $111.4 billion. That pattern is not new and it is not a crisis. It is a decade-old rotation in which the marginal buyer of American debt stopped being a foreign government and became a foreign asset manager.

Foreign holdings of federal debt were around $9.2 trillion at the end of 2025, up roughly $1.5 trillion in four years. Japan still holds about $1.2 trillion, the UK about $0.9 trillion, China about $0.7 trillion. Nobody has run.

The distinction matters because the two buyers behave differently. A central bank holds Treasuries because it needs dollar reserves and has nowhere else to put them at size. It is insensitive to price and it rarely leaves. An asset manager holds them because the yield beats the alternative this quarter. It is very sensitive to price, and it demands to be paid for duration.

Three stories carried most of the de-dollarisation traffic this month. All three are real events. None of them says what the summary said.

The dollar’s own scoreboard agrees. Its share of allocated global reserves rose to 57.13% in the first quarter of 2026, from 56.42% at the end of 2025. The low was 56.9% in the third quarter of 2025, the weakest since 1995, and the peak was 72% in 2001. So the long slide is real and the short-term direction this year is up. Anyone telling you the dollar share just hit a record low is quoting last year’s number.

This is the part worth being precise about, because the honest version is more useful than the dramatic one. There is no run on the Treasury market. There is a slow, twenty-year reallocation by official holders, a fast reallocation by everyone else toward gold, and a private sector that still buys dollars at a price. The question is what that price has to be.

The pressure is not coming from abroad. It is coming from the coupon.

The Congressional Budget Office puts federal interest costs at $1,039 billion in fiscal 2026. That is 3.3% of GDP, above the 1991 record, and 18.6% of federal revenue, also a record. Interest is now the second-largest line in the budget, behind Social Security and ahead of defence and Medicare. Every time maturing debt rolls over at a higher coupon than the one it replaces, that line grows without a single new spending decision being taken.

And the long end is not cooperating. The 10-year closed at 4.84% on 9 September, its highest since October 2023, and was near 4.95% by the 11th, with the 30-year at 5.36%. Thirty-year mortgages have been printing between 6.8% and 7.1%. The policy rate is 3.50% to 3.75%, and futures put roughly two-in-three odds on Kevin Warsh’s Fed raising it on 16 September, with the chair saying underlying inflation has not “meaningfully improved” and at least one governor inclined to hold.

So the Treasury faces the worst version of the problem. Short rates that may go up, long rates it cannot talk down, and a fuel complex doing the opposite of what disinflation needs. On-highway diesel hit $5.967 a gallon on 7 September, above the June 2022 record. The Strategic Petroleum Reserve is at 285.4 million barrels, the lowest since November 1982, after 172 million barrels were released when Iran choked off traffic through the Strait of Hormuz. Shanghai crude has traded above $121 against Brent near $100, because China has come back into the market to refill its own 1.4 billion barrel reserve and is outbidding everyone for the cargoes.

We took the machinery apart three weeks ago, so this is the short version. The Treasury has been buying back long-dated bonds and paying for them with short-dated bills, a trade Scott Bessent named the Treasury Twist. It built its cash balance at the Fed to roughly $950 billion to fund the operation. And the GENIUS Act quietly turned every payment stablecoin into a buyer of Treasury bills, because its reserve rules permit almost nothing else.

There is one piece we did not cover then, and it is the most elegant part. Bessent has been pressing the Fed to widen its FIMA repo facility, the window where a foreign central bank posts Treasuries and takes dollars back. Japan deposits the bonds, gets the dollars it needs to buy yen, and keeps the bonds on its balance sheet. No sale prints. Nothing shows up in the holdings data as a foreign government leaving the Treasury market, because in the accounting sense none has. It is a pawnshop, and the gap between pawning and selling is the entire point of the facility.

That earlier piece ended with three things to watch. The first was whether the buyback held. It has now been tested, and before we get to the result, two numbers being quoted around it need pinning down.

The first is the size of the legislated bid, where the figures in circulation contradict each other. A stablecoin market of roughly $270 billion implies about $125 billion of Treasury bills held in today’s proportions, against some $6 trillion outstanding, so under 2%. Tether on its own reports $141 billion of Treasury exposure and ranks around eighteenth among all holders of US government debt, ahead of Germany and the UAE. Both get quoted as though they measure the same thing. They don’t. Tether’s number counts indirect exposure through repo and money market funds alongside bills held outright, so it cannot be stacked on the market-wide estimate, and you should be wary of anyone who does. Either way the order of magnitude holds, and the Kansas City Fed makes the sharper point anyway: money entering stablecoins largely leaves deposits and money funds that were buying government paper already.

The second is wrong outright, and worth killing because it sends people looking in the wrong place. The Federal Reserve is not printing money to fund the buybacks. Quantitative tightening ended on 1 December 2025. The Fed now rolls maturing Treasuries at auction and reinvests agency principal into Treasury bills, holding its balance sheet roughly flat rather than growing it. The repurchases are financed from the Treasury’s own cash and from bill issuance. If you are waiting for a visible jump in the Fed’s balance sheet as your signal, you will be watching a chart where nothing happens while the maturity structure of the debt changes underneath you.

The useful thing about a policy this specific is that it comes with a test date. That test has already been run.

On 9 September, the day the Treasury confirmed a repurchase of up to $6 billion, the ten-year yield closed at 4.84%, the highest in nearly three years. It carried on to roughly 4.95% two days later, with the thirty-year at 5.36%. An operation designed to support long-end prices was followed within hours by long-end prices falling. The analyst read was unsentimental: a few billion dollars of repurchases does not offset a widening deficit, sticky inflation, and heavy government issuance everywhere else in the world at the same time.

Sit with that for a moment, because it is the closest thing to evidence this argument has produced. Not a forecast about what the bond market might do to a government defending its own debt. Something it did, on a known date, in response to a known announcement.

There is a playbook for this and it was never secret. The United States ran it from the 1940s, holding yields below the rate of inflation until debt taken on for a war shrank against a growing economy. Japan has run its own version since the 1990s and now carries government debt somewhere between 188% and 204% of GDP depending on which measure you take, financed overwhelmingly at home. Neither country defaulted. Both moved the cost onto whoever was holding cash and bonds. That is what financial repression is for, and that is who pays for it.

If the demand cannot be manufactured fast enough, the adjustment shows up somewhere else. It usually shows up in the currency and in what the patient money does instead.

Central banks bought a net 289 tonnes of gold in the second quarter, up 62% year on year and the strongest second quarter on record, and they did it while the gold price was posting its steepest quarterly fall in a decade. That is the signature of a reserve manager buying to a policy rather than to a chart. Gold ETFs took $18 billion in August, the second-largest month on record, with holdings at an all-time high of 4,189 tonnes.

Capital is being called home, too, and not only from Washington. All 27 EU member states have agreed a Savings and Investments Union, after Ursula von der Leyen described Europe’s €10 trillion of bank deposits as lazy and put a number on the leakage: roughly €300 billion a year leaving the continent, mostly for the United States. China imposed a 20% tax on offshore trusts in July, with a transitional window that closes on 22 October and three years of retroactive cover, while Hong Kong banks restrict mainland clients from buying overseas stocks. Japan’s ten-year yield reached 3% on 1 September, the highest since 1996, which for the first time in a generation gives a Japanese institution a reason to fund at home.

None of those is aimed at the Treasury market. All of them take a slice out of the flow that used to end up there by default.

We would rather hand you a calendar than a forecast.

The one to weight most is the least dramatic. If the doubled buyback authority is quietly renewed on 4 November and the bill share keeps climbing, the duration shift is policy rather than a quarter-end convenience, and it will keep going until something forces it to stop.

Take the bear case seriously and it still does not get you to a collapse. The counter-evidence is strong and we would rather put it in the article than in a footnote. Private foreign demand is robust. The dollar’s reserve share went up this year. Norway is cutting Treasuries and staying in dollars. A 30-year yield of 5.36% is the highest compensation for lending to the US government since before the financial crisis, and for an investor with a matching thirty-year liability it may be a perfectly good trade.

Our reading is narrower. The United States is not losing access to credit. It is losing the ability to set the price of long credit, and it is responding by moving the debt to where the price is easier to control and by writing new holders into statute. The 9 September reaction is the part that should worry a holder of long-dated dollar debt more than any of the de-dollarisation headlines. A government announced that it would buy its own bonds, and the price of those bonds went down. That is a slower, more administrative process than a funding crisis, and it is harder to hedge precisely because nothing breaks on any given morning.

What follows from it for anyone holding assets is unglamorous. If you are being paid a nominal yield in the currency doing the repressing, read the real return, not the coupon. If the 30-year pays 5.36% and diesel just set a record, the question is not whether the yield is high by history. It is whether you are being paid enough for a decade of policy you do not control.

We hold real assets, income in more than one currency, and a deliberate spread of jurisdictions, and gold does a specific job in that mix rather than a decorative one. That is a disclosed interest, not a recommendation, and the bull case above is the honest argument against us.

The general form is the one we keep returning to. Governments under fiscal pressure do not usually announce that they are changing the deal. They adjust the rules about where money has to sit, who has to hold what, and on what terms it may leave. The last two years gave us Indonesia doing it to its residents, Europe doing it to its savers, China doing it to its trusts, and now Washington doing it to the maturity structure of the biggest bond market on earth. Read the regulation, not the speech.

This analysis is for informational purposes only and is not personal investment advice. Valid as of publication date; conditions evolve. Past returns are not indicative of future results. Pressure-test the framing against your own thesis before acting on it.